This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Single-delete Nim

Masato Shinoda Division of Natural Sciences, Nara Women’s University. [email protected]
Abstract

The classic game of Nim has been well-known for many years, inspiring numerous variations. One such variant is Delete Nim, where players take turns eliminating one pile of stones and splitting the remaining pile into two smaller piles. In this paper we generalize the game to include the case of nn piles. On each turn, a player eliminates one pile and splits one of the remaining piles into two smaller piles. We specifically analyze the case where n=4n=4, deriving the conditions for a winning strategy.

1 Introduction and Main Result

1.1 Introduction

Nim is a well-known game that has been studied extensively over the years. In the most popular version of Nim, there are three piles of stones, and two players take turns. On their turn, a player selects one pile and removes any positive number of stones from that pile. The players continue taking turns, and the player who removes the last stone wins the game. This game can also be played with more than three piles following the same rules.

Nim is a two-player, zero-sum, finite, deterministic, and perfect information game with no possibility of a draw. Therefore, every configuration can be classified as either an N-position, where the player whose turn it is can force a win, or a P-position, where the player whose turn it is will inevitably lose if the opponent plays optimally. The classification of N-positions and P-positions in the aforementioned version of Nim was demonstrated by Bouton [3] in 1902. It became known that the conditions for the winning strategy in this game include mathematically intriguing structures.

Several variants of Nim have been proposed by modifying the rules of legal moves, such as Moore’s game [4] and Wythoff’s game [7]. The winning strategies of these variants have been studied extensively, and their mathematical formulations have attracted significant interest. In this study, we generalize Delete Nim, a game introduced by Abuku and Suetsugu [2], where players eliminate one pile and then split one of the remaining piles into two smaller piles in one move. We then investigate the winning strategy for this generalized version.

1.2 Delete Nim

In this section, we first describe the rules of Delete Nim, which form the basis of this study. There are two variations of the Delete Nim rules provided in [2] and [6]. Although the rules differ in presentation, they are equivalent with respect to legal moves. In this paper, we adopt the latter rules111In [2], the rule described here is referred to as the “Variant of Delete Nim” (VDN)..

In Delete Nim, there are two piles of stones. On each turn, players perform the following two actions sequentially (we refer to these combined actions as a move), then pass the turn to their opponent.

  • Select one of the piles and remove all the stones from it, eliminating the pile.

  • Split the remaining pile into two new piles, ensuring that each new pile contains at least one stone.

A player unable to make this move on their turn loses the game. We denote a position in Delete Nim by y,z\langle y,z\rangle, where yy and zz represent the number of stones in the two piles. Let \mathbb{N} be the set of positive integers. Then, the set of all possible game positions (without specifying the current player’s turn) is denoted by G2={y,z|y,z}G_{2}=\{\,\langle y,z\rangle\,|\,y,z\in\mathbb{N}\}. At the position 1,1\langle 1,1\rangle, the player whose turn it loses the game.

Similar to Nim, Delete Nim is a two-player, zero-sum, perfect information, deterministic, finite game with no possibility of a draw. Therefore, each position in G2G_{2} can be classified as either an N-position or a P-position. The condition for determining this classification is as follows.

Theorem 1.1 ([2, 6]).

In Delete Nim, a position y,z\langle y,z\rangle is a P-position if both yy and zz are odd; otherwise, it is an N-position.

This theorem is included in Proposition 4.1 in this paper. It should be noted that in [2], not only is the above classification provided, but the Sprague-Grundy number of each position is also determined. In an N-position, a player can ensure winning by forcing their opponent into P-positions with optimal moves. For example, when the current position is 4,3\langle 4,3\rangle, the player on turn can proceed as 4,33,1\langle 4,3\rangle\to\langle 3,1\rangle (and then the game proceeds as 2,11,1\to\langle 2,1\rangle\to\langle 1,1\rangle and ends) and secure winning. Therefore, identifying the P-positions is equivalent to understanding a winning strategy and knowing the conditions for winning.

1.3 Single-delete Nim

In this section, we define an extended rule that generalizes Delete Nim to any number of piles. Let nn be an integer greater than or equal to 2.

Definition 1.2 (Single-delete Nim).

There are nn piles of stones. On each turn, the two players perform the following two actions sequentially, then pass the turn to their opponent.

  • Select one of the piles and remove all the stones from it, eliminating the pile.

  • Select one of the remaining n1n-1 piles and split it into two new piles, ensuring that each new pile contains at least one stone.

A player unable to make this move on their turn loses the game.

In Single-delete Nim, it is important to note that the number of piles remains nn at the end of each turn.

Remark 1.3.

We introduce the term “Single-delete Nim” in this paper. When generalizing the number of piles in Delete Nim, various extended rules can be considered depending on how piles are eliminated and split. The name “Single-delete Nim” explicitly refers to the rule of eliminating only one pile. For more details, see Abuku et al. [1].

The following result for Single-delete Nim with three piles is known from Sakai [5]. To present the winning and losing conditions, we define the following notation: For zz\in\mathbb{N}, if zz is divisible by 2k2^{k} but not divisible by 2k+12^{k+1}, we denote it as v2(z)=kv_{2}(z)=k. When zz is odd, we define v2(z)=0v_{2}(z)=0.

Theorem 1.4.

A necessary and sufficient condition for the position x,y,z\langle x,y,z\rangle in Single-delete Nim with three piles to be a P-position is

(\bigstar)     v2(x)=v2(y)=v2(z)v_{2}(x)=v_{2}(y)=v_{2}(z).

A proof of Theorem 1.4 is provided in [5] (pp.59-63). However, as it is also relevant for the case of four piles, we restate it in Section 2.1 of this paper.

1.4 Main Result

As the main theorem of this paper, we present the necessary and sufficient condition to be a P-position in Single-delete Nim with four piles as Theorem 1.5. To state this theorem, we introduce notation related to the binary representation of positive integers. Let Ik(z)I_{k}(z) represent the kk-th digit from the right in the binary representation of zz\in\mathbb{N}. In other words, if the quotient of zz divided by 2k12^{k-1} is even, then Ik(z)=0I_{k}(z)=0; if the quotient is odd, then Ik(z)=1I_{k}(z)=1. We note that v2(z)=min{k|Ik(z)=1}1v_{2}(z)=\min\{\,k\,|\,I_{k}(z)=1\}-1.

Theorem 1.5.

In Single-delete Nim with four piles, for a position w,x,y,z\langle w,x,y,z\rangle, let a=v2(w),b=v2(x),c=v2(y),d=v2(z)a=v_{2}(w),\,b=v_{2}(x),\,c=v_{2}(y),\,d=v_{2}(z). When abcda\leq b\leq c\leq d, the necessary and sufficient condition for w,x,y,z\langle w,x,y,z\rangle to be a P-position is that one of the following conditions (1) to (5) is satisfied.

  1. (1).

    a=b=c=da=b=c=d.

  2. (2).

    a<b=c=da<b=c=d and
    (2A)  Ib+1(w)=0I_{b+1}(w)=0.

  3. (3).

    a<b<c=da<b<c=d and all of (3A)-(3C) are satisfied.
    (3A)  Ic+1(w)=Ic+1(x)=0I_{c+1}(w)=I_{c+1}(x)=0,
    (3B)  Ik(w)+Ik(x)1I_{k}(w)+I_{k}(x)\geq 1 for b+2kcb+2\leq k\leq c,
    (3C)  Ib+1(w)=1I_{b+1}(w)=1.

  4. (4).

    a<b<c<da<b<c<d and all of (4A)-(4E) are satisfied.
    (4A)  Id+1(w)=Id+1(x)=Id+1(y)=0I_{d+1}(w)=I_{d+1}(x)=I_{d+1}(y)=0,
    (4B)  Ij(w)+Ij(x)+Ij(y)2I_{j}(w)+I_{j}(x)+I_{j}(y)\geq 2 for c+2jdc+2\leq j\leq d,
    (4C)  Ic+1(w)=Ic+1(x)=1I_{c+1}(w)=I_{c+1}(x)=1,
    (4D)  Ik(w)+Ik(x)1I_{k}(w)+I_{k}(x)\geq 1 for b+2kcb+2\leq k\leq c,
    (4E)  Ib+1(w)=1I_{b+1}(w)=1.

  5. (5).

    a<b<c<da<b<c<d and all of (5A)-(5F) are satisfied.
    (5A)  Ii(w)+Ii(x)+Ii(y)+Ii(z){0,3,4}I_{i}(w)+I_{i}(x)+I_{i}(y)+I_{i}(z)\in\{0,3,4\} for id+2i\geq d+2,
    (5B)  Id+1(w)=Id+1(x)=Id+1(y)=1I_{d+1}(w)=I_{d+1}(x)=I_{d+1}(y)=1,
    (5C)  Ij(w)+Ij(x)+Ij(y)2I_{j}(w)+I_{j}(x)+I_{j}(y)\geq 2 for c+2jdc+2\leq j\leq d,
    (5D)  Ic+1(w)=Ic+1(x)=1I_{c+1}(w)=I_{c+1}(x)=1,
    (5E)   Ik(w)+Ik(x)1I_{k}(w)+I_{k}(x)\geq 1 for b+2kcb+2\leq k\leq c,
    (5F)  Ib+1(w)=1I_{b+1}(w)=1.

The organization of this paper is as follows. In Section 2, we present a proof of Theorem 1.4, which provides the classification of N-positions and P-positions for Single-delete Nim with three piles, along with several useful propositions. In Section 3, we give a proof of Theorem 1.5, which addresses the case with four piles. In Section 4, we offer a brief analysis of the scenario when the number of piles is five or more.

2 Single-delete Nim with three piles

In this section, we address a proof of Theorem 1.4 for the case with three piles. To support this proof, we prepare the following propositions. These are nearly self-evident facts, so the proofs are omitted here, but we present them collectively as they will be frequently used in the next chapter as well.

Proposition 2.1.

For x,y,zx,y,z\in\mathbb{N} satisfying x+y=zx+y=z, the following facts (i) and (ii) hold.

  1. (i).

    If v2(x)=v2(y)v_{2}(x)=v_{2}(y), then v2(z)>v2(x)v_{2}(z)>v_{2}(x),

  2. (ii).

    If v2(x)v2(y)v_{2}(x)\neq v_{2}(y), then v2(z)=min{v2(x),v2(y)}v_{2}(z)=\min\{v_{2}(x),v_{2}(y)\}.

Proposition 2.2.

When zz\in\mathbb{N} and v2(z)1v_{2}(z)\geq 1, for any non-negative integer k<v2(z)k<v_{2}(z), there exist x,yx,y\in\mathbb{N} such that x+y=zx+y=z and v2(x)=v2(y)=kv_{2}(x)=v_{2}(y)=k.

Proof of Theorem 1.4. Let PP denote the set of all positions that satisfy condition (\bigstar). This set is a subset of G3={x,y,z|x,y,z}G_{3}=\{\,\langle x,y,z\rangle\,|\,x,y,z\in{\mathbb{N}}\}, which represents all positions in Single-delete Nim with three piles. Let NN denote the set of all other positions. In the progression of this game, the total number of stones across all piles decreases monotonically. Since PP includes the terminal position 1,1,1\langle 1,1,1\rangle, it is sufficient to show that no position in PP can be moved to any another position in PP, and that any position in NN can be moved to some position in PP.

First, assume x,y,zP\langle x,y,z\rangle\in P. In this case, since v2(x)=v2(y)=v2(z)v_{2}(x)=v_{2}(y)=v_{2}(z), it is sufficient to consider the next move where the pile with zz stones is eliminated and the pile with xx stones is split into two piles with xx^{\prime} and x′′x^{\prime\prime} stones, resulting in the position x,x′′,y\langle x^{\prime},x^{\prime\prime},y\rangle. By Proposition 2.1(i), if v2(x)=v2(x′′)v_{2}(x^{\prime})=v_{2}(x^{\prime\prime}) then v2(x)<v2(x)=v2(y)v_{2}(x^{\prime})<v_{2}(x)=v_{2}(y). This implies that x,x′′,y\langle x^{\prime},x^{\prime\prime},y\rangle cannot satisfy (\bigstar). Therefore, any position in PP cannot be moved to another position in PP.

Next, assume x,y,zN\langle x,y,z\rangle\in N. Since v2(x)=v2(y)=v2(z)v_{2}(x)=v_{2}(y)=v_{2}(z) is not satisfied, it is sufficient to consider the case where v2(x)<v2(y)v_{2}(x)<v_{2}(y). By Proposition 2.2, the pile with yy stones can be split into two piles with yy^{\prime} and y′′y^{\prime\prime} stones such that v2(y)=v2(y′′)=v2(x)v_{2}(y^{\prime})=v_{2}(y^{\prime\prime})=v_{2}(x). Then, by eliminating the pile with zz stones and splitting, the position can be moved to x,y,y′′P\langle x,y^{\prime},y^{\prime\prime}\rangle\in P. Thus, any position in NN can be moved to a position in PP. \Box

3 Single-delete Nim with four piles

3.1 Necessity of given conditions

In this section, we present a detailed proof of Theorem 1.5, which establishes the necessary and sufficient conditions for a position in Single-delete Nim with four piles to qualify as a P-position. We first briefly explain the necessity of some conditions (1) to (5) in Theorem 1.5 through a comparison of specific examples. The necessity of all the conditions will be discussed in detail in the proof of Lemma 3.2 later.

  1. (1).

    A specific example of a position in this case is w,x,y,z=1440,864,\langle w,x,y,z\rangle=\langle 1440,864,
    672,1120672,1120\rangle. In its binary representation, it is

    w10110100000x01101100000y01010100000z10001100000,\begin{array}[]{ll}w&10110100000\\ x&01101100000\\ y&01010100000\\ z&10001100000,\end{array}

    where the rightmost 1’s are in the same digits.

  2. (2).

    A specific example of a position in this case is w,x,y,z=294,208,\langle w,x,y,z\rangle=\langle 294,208,
    304,432304,432\rangle. In its binary representation, it is

    w100100110x011010000y100110000z110110000.\begin{array}[]{ll}w&100100110\\ x&011010000\\ y&100110000\\ z&110110000.\end{array}

    In a similar position

    w10011¯0110x011010000y100110000z110110000\begin{array}[]{ll}w&1001\underline{1}0110\\ x&011010000\\ y&100110000\\ z&110110000\end{array}

    where condition (2A) is not satisfied, by eliminating xx and partitioning ww into ww^{\prime} and w′′w^{\prime\prime}, it is possible to fulfill condition (2) in a move, resulting in

    w100100110w′′000010000y100110000z110110000.\begin{array}[]{ll}w^{\prime}&100100110\\ w^{\prime\prime}&000010000\\ y&100110000\\ z&110110000.\end{array}
  3. (3).

    A specific example of a position in this case is w,x,y,z=669,468,\langle w,x,y,z\rangle=\langle 669,468,
    800,288800,288\rangle. In its binary representation, it is

    w1010011101x0111010100y1100100000z0100100000.\begin{array}[]{ll}w&1010011101\\ x&0111010100\\ y&1100100000\\ z&0100100000.\end{array}

    In a similar position

    w101000¯1101x011100¯0100y1100100000z0100100000\begin{array}[]{ll}w&10100\underline{0}1101\\ x&01110\underline{0}0100\\ y&1100100000\\ z&0100100000\end{array}

    where condition (3B) is not satisfied, by eliminating zz and partitioning yy into yy^{\prime} and y′′y^{\prime\prime}, it is possible to fulfill condition (3) in a move, resulting in

    w1010001101x0111000100y1100010000y′′0000010000.\begin{array}[]{ll}w&1010001101\\ x&0111000100\\ y^{\prime}&1100010000\\ y^{\prime\prime}&0000010000.\end{array}
  4. (4).

    A specific example of a position in this case is w,x,y,z=11133,\langle w,x,y,z\rangle=\langle 11133,
    12716,7136,1331212716,7136,13312\rangle. In its binary representation, it is

    w10101101111101x11000110101100y01101111100000z11010000000000.\begin{array}[]{ll}w&10101101111101\\ x&11000110101100\\ y&01101111100000\\ z&11010000000000.\end{array}

    In a similar position

    w10101101111101x11000110101100y0110110¯1100000z11010000000000\begin{array}[]{ll}w&10101101111101\\ x&11000110101100\\ y&011011\underline{0}1100000\\ z&11010000000000\end{array}

    where condition (4B) is not satisfied, by eliminating xx and partitioning zz into zz^{\prime} and z′′z^{\prime\prime}, it is possible to fulfill condition (3) in a move, resulting in

    w10101101111101y01101101100000z11001110000000z′′00000010000000.\begin{array}[]{ll}w&10101101111101\\ y&01101101100000\\ z^{\prime}&11001110000000\\ z^{\prime\prime}&00000010000000.\end{array}
  5. (5).

    A specific example of a position in this case is w,x,y,z=45053,\langle w,x,y,z\rangle=\langle 45053,
    62932,32576,6451262932,32576,64512\rangle. In its binary representation, it is

    w1010111111111101x1111010111010100y0111111101000000z1111110000000000.\begin{array}[]{ll}w&1010111111111101\\ x&1111010111010100\\ y&0111111101000000\\ z&1111110000000000.\end{array}

    In a similar position

    w1010111111111101x1111010111010100y0110¯111101000000z1111110000000000\begin{array}[]{ll}w&1010111111111101\\ x&1111010111010100\\ y&011\underline{0}111101000000\\ z&1111110000000000\end{array}

    where condition (5A) is not satisfied, by eliminating xx and partitioning zz into zz^{\prime} and z′′z^{\prime\prime}, it is possible to fulfill condition (4) in a move, resulting in

    w1010111111111101y0110111101000000z1110110000000000z′′0001000000000000.\begin{array}[]{ll}w&1010111111111101\\ y&0110111101000000\\ z^{\prime}&1110110000000000\\ z^{\prime\prime}&0001000000000000.\end{array}

3.2 Proof of Theorem 1.5

In this section, we will prove Theorem 1.5. Let G4={w,x,y,z|w,x,y,zG_{4}=\{\,\langle w,x,y,z\rangle\,|\,w,x,y,z
}\in\mathbb{N}\} be the set of all positions in Single-delete Nim with four piles. We say that w,x,y,zG4\langle w,x,y,z\rangle\in G_{4} is a standard form if it satisfies v2(w)v2(x)v2(y)v2(z)v_{2}(w)\leq v_{2}(x)\leq v_{2}(y)\leq v_{2}(z). Rearranging the elements of w,x,y,zG4\langle w,x,y,z\rangle\in G_{4} in this order is referred to as the standardization of w,x,y,z\langle w,x,y,z\rangle. We define the subset PG4P\subset G_{4} such that a position w,x,y,z\langle w,x,y,z\rangle belongs to PP if and only if its standardization satisfies any of the conditions (1) to (5) in Theorem 1.5. Let NN denote the set of all other positions in G4G_{4}. In other words, PP and NN correspond to the sets of P-positions and N-positions for this game, respectively.

In the progression of this game, the total sum of stones across all piles decreases monotonically, and since PP includes the terminal position 1,1,1,1\langle 1,1,1,1\rangle, the proof of Theorem 1.5 requires demonstrating the following two lemmas.

Lemma 3.1.

If w,x,y,zP\langle w,x,y,z\rangle\in P, then it cannot be moved to any position in PP.

Lemma 3.2.

If w,x,y,zN\langle w,x,y,z\rangle\in N, then it can be moved to some position in PP.

To prove these lemmas, we first define common notation. If the standard form of w,x,y,z\langle w,x,y,z\rangle satisfies condition (ii) of Theorem 1.5 (1i51\leq i\leq 5), we denote w,x,y,zPi\langle w,x,y,z\rangle\in P_{i} and define the sets P1,P2,P3,P4,P5P_{1},P_{2},P_{3},P_{4},P_{5} accordingly. Thus, P=i=15Pi\displaystyle P=\bigcup_{i=1}^{5}P_{i}, and by conditions (1) to (5) we have PiPj=P_{i}\cap P_{j}=\emptyset for any iji\neq j.

We now present the following proposition to support our proof of the two lemmas.

Proposition 3.3.

If w,xw,x\in\mathbb{N} satisfy Ib+1(w)=Ib+1(x)=1I_{b+1}(w)=I_{b+1}(x)=1 and Ic+1(w)=Ic+1(x)I_{c+1}(w)=I_{c+1}(x) (b<cb<c), and additionally satisfy Ij(w)+Ij(x)1I_{j}(w)+I_{j}(x)\geq 1 for all jj with b+2jcb+2\leq j\leq c, then Ic+1(w+x)=1I_{c+1}(w+x)=1.

Considering the carry in binary addition, it is clear that the above proposition holds. With this in mind, we will now proceed to prove Lemma 3.1 and Lemma 3.2.

Proof of Lemma 3.1. For a position w,x,y,z\langle w,x,y,z\rangle where a=v2(w),b=v2(x),c=v2(y),d=v2(z)a=v_{2}(w),b=v_{2}(x),c=v_{2}(y),d=v_{2}(z), if any of the following conditions hold:

(\trianglea=b=c<da=b=c<d, a=b<c=da=b<c=d, a=b<c<da=b<c<d, a<b=c<da<b=c<d

then none of the conditions (1) to (5) can be satisfied, and thus the position is not included in PP. This fact will be frequently used in the explanation below. In this proof, when transitioning from w,x,y,z\langle w,x,y,z\rangle to w,x,y,z\langle w^{\prime},x^{\prime},y^{\prime},z^{\prime}\rangle as a move, we show that if w,x,y,zP\langle w^{\prime},x^{\prime},y^{\prime},z^{\prime}\rangle\in P then w,x,y,zP\langle w,x,y,z\rangle\not\in P. We consider the reverse of the operation:

  • First, select two of w,x,y,zw^{\prime},x^{\prime},y^{\prime},z^{\prime} and combine their sum into a single number.

  • Then, add a new element uu^{\prime}\in\mathbb{N}.

We consider each case where w,x,y,zPi\langle w^{\prime},x^{\prime},y^{\prime},z^{\prime}\rangle\in P_{i} for 1i51\leq i\leq 5.

  1. a).

    In case w,x,y,zP1\langle w^{\prime},x^{\prime},y^{\prime},z^{\prime}\rangle\in P_{1}:
    Since v2(w)=v2(x)=v2(y)=v2(z)v_{2}(w^{\prime})=v_{2}(x^{\prime})=v_{2}(y^{\prime})=v_{2}(z^{\prime}), it is sufficient to consider only the case where the sum y+zy^{\prime}+z^{\prime} is formed in the reverse operation. In this situation, by Proposition 2.1(i), we have v2(w)=v2(x)<v2(y+z)v_{2}(w^{\prime})=v_{2}(x^{\prime})<v_{2}(y^{\prime}+z^{\prime}). Therefore, regardless of the element uu^{\prime} added, u,w,x,y+z\langle u^{\prime},w^{\prime},x^{\prime},y^{\prime}+z^{\prime}\rangle corresponds to (\triangle), and cannot be included in any PiP_{i}.

  2. b).

    In case w,x,y,zP2\langle w^{\prime},x^{\prime},y^{\prime},z^{\prime}\rangle\in P_{2}:
    Suppose v2(w)<v2(x)=v2(y)=v2(z)v_{2}(w^{\prime})<v_{2}(x^{\prime})=v_{2}(y^{\prime})=v_{2}(z^{\prime}). Let b=v2(x)b=v_{2}(x^{\prime}), and we will use the following facts:
    (1\clubsuit_{1}) From (2A), we have Ib+1(w)=0I_{b+1}(w^{\prime})=0.
    (2\clubsuit_{2}) Ib+1(w+z)=1I_{b+1}(w^{\prime}+z^{\prime})=1?D

    • We consider whether u,w,x,y+z\langle u^{\prime},w^{\prime},x^{\prime},y^{\prime}+z^{\prime}\rangle, formed by the reverse operation where the sum y+zy^{\prime}+z^{\prime} is taken and uu^{\prime} is added, is included in PP. Here, we have v2(w)<v2(x)<v2(y+z)v_{2}(w^{\prime})<v_{2}(x^{\prime})<v_{2}(y^{\prime}+z^{\prime}).

      • If v2(u)>v2(x)v_{2}(u^{\prime})>v_{2}(x^{\prime}), then by (1\clubsuit_{1}), it cannot satisfy (3C), (4E) or (5F).

      • If v2(u)=v2(x)v_{2}(u^{\prime})=v_{2}(x^{\prime}) or v2(u)=v2(w)v_{2}(u^{\prime})=v_{2}(w^{\prime}), it corresponds to (\triangle).

      • If v2(u)<v2(x)v_{2}(u^{\prime})<v_{2}(x^{\prime}) and v2(u)v2(w)v_{2}(u^{\prime})\neq v_{2}(w^{\prime}), then by (1\clubsuit_{1}), it cannot satisfy (4C) or (5D).

      Therefore, we can conclude that u,w,x,y+zP\langle u^{\prime},w^{\prime},x^{\prime},y^{\prime}+z^{\prime}\rangle\not\in P.

    • We consider whether u,w+z,x,y\langle u^{\prime},w^{\prime}+z^{\prime},x^{\prime},y^{\prime}\rangle, formed by the reverse operation where the sum w+zw^{\prime}+z^{\prime} is taken and uu^{\prime} is added, is included in PP. Here, we have v2(w+z)<v2(x)=v2(y)v_{2}(w^{\prime}+z^{\prime})<v_{2}(x^{\prime})=v_{2}(y^{\prime}).

      • If v2(u)>v2(y)v_{2}(u^{\prime})>v_{2}(y^{\prime}), it corresponds to (\triangle).

      • If v2(u)v2(y)v_{2}(u^{\prime})\leq v_{2}(y^{\prime}), then by (2\clubsuit_{2}), it cannot satisfy (2A) or (3A).

      Therefore, we can conclude that u,w+z,x,yP\langle u^{\prime},w^{\prime}+z^{\prime},x^{\prime},y^{\prime}\rangle\not\in P.

  3. c).

    In case w,x,y,zP3\langle w^{\prime},x^{\prime},y^{\prime},z^{\prime}\rangle\in P_{3}:
    Suppose v2(w)<v2(x)<v2(y)=v2(z)v_{2}(w^{\prime})<v_{2}(x^{\prime})<v_{2}(y^{\prime})=v_{2}(z^{\prime}). Let c=v2(y)c=v_{2}(y^{\prime}), and we will use the following facts:
    (1\diamondsuit_{1}) From (3A), we have Ic+1(w)=Ic+1(x)=0I_{c+1}(w^{\prime})=I_{c+1}(x^{\prime})=0.
    (2\diamondsuit_{2}) Ic+1(x+z)=1I_{c+1}(x^{\prime}+z^{\prime})=1.
    (3\diamondsuit_{3}) From (3B) and Proposition 3.3, we have Ic+1(w+x)=1I_{c+1}(w^{\prime}+x^{\prime})=1.

    • We consider whether u,w,x,y+z\langle u^{\prime},w^{\prime},x^{\prime},y^{\prime}+z^{\prime}\rangle, formed by the reverse operation where the sum y+zy^{\prime}+z^{\prime} is taken and uu^{\prime} is added, is included in PP. Here, we have v2(w)<v2(x)<c<v2(y+z)v_{2}(w^{\prime})<v_{2}(x^{\prime})<c<v_{2}(y^{\prime}+z^{\prime}).

      • If v2(u)>cv_{2}(u^{\prime})>c, then by (1\diamondsuit_{1}), it cannot satisfy (3B), (4D) or (5E).

      • If v2(u)=cv_{2}(u^{\prime})=c, then by (1\diamondsuit_{1}), it cannot satisfy (4C) or (5D).

      • If v2(u)<cv_{2}(u^{\prime})<c, v2(u)v2(w)v_{2}(u^{\prime})\neq v_{2}(w^{\prime}) and v2(u)v2(x)v_{2}(u^{\prime})\neq v_{2}(x^{\prime}), then by (1\diamondsuit_{1}), it cannot satisfy (4B) or (5C).

      • If v2(u)=v2(w)v_{2}(u^{\prime})=v_{2}(w^{\prime}) or v2(u)=v2(x)v_{2}(u^{\prime})=v_{2}(x^{\prime}), it corresponds to (\triangle).

      Therefore, we can conclude that u,w,x,y+zP\langle u^{\prime},w^{\prime},x^{\prime},y^{\prime}+z^{\prime}\rangle\not\in P.

    • We consider whether u,w,x+z,y\langle u^{\prime},w^{\prime},x^{\prime}+z^{\prime},y^{\prime}\rangle, formed by the reverse operation where the sum x+zx^{\prime}+z^{\prime} is taken and uu^{\prime} is added, is included in PP. Here, we have v2(w)<v2(x+z)<v2(y)v_{2}(w^{\prime})<v_{2}(x^{\prime}+z^{\prime})<v_{2}(y^{\prime}).

      • If v2(u)>v2(y)v_{2}(u^{\prime})>v_{2}(y^{\prime}), then by (1\diamondsuit_{1}), it cannot satisfy (4C) or (5D).

      • If v2(u)=v2(y)v_{2}(u^{\prime})=v_{2}(y^{\prime}), then by (2\diamondsuit_{2}), it cannot satisfy (3A).

      • If v2(u)<v2(y)v_{2}(u^{\prime})<v_{2}(y^{\prime}), v2(u)v2(x+z)v_{2}(u^{\prime})\neq v_{2}(x^{\prime}+z^{\prime}) and v2(u)v2(w)v_{2}(u^{\prime})\neq v_{2}(w^{\prime}), then by (1\diamondsuit_{1}) and (2\diamondsuit_{2}), it cannot satisfy (4A) or (5B).

      • If v2(u)=v2(x+w)v_{2}(u^{\prime})=v_{2}(x^{\prime}+w^{\prime}) or v2(u)=v2(z)v_{2}(u^{\prime})=v_{2}(z^{\prime}), it corresponds to (\triangle).

      Therefore, we can conclude that u,w,x+z,yP\langle u^{\prime},w^{\prime},x^{\prime}+z^{\prime},y^{\prime}\rangle\not\in P.

    • For u,w+z,x,y\langle u^{\prime},w^{\prime}+z^{\prime},x^{\prime},y^{\prime}\rangle, formed by the reverse operation where the sum w+zw^{\prime}+z^{\prime} is taken and uu^{\prime} is added, it can similarly be shown that it is not included in PP, just as in the case where the sum x+zx^{\prime}+z^{\prime} is taken and uu^{\prime} is added.

    • We consider whether u,w+x,y,z\langle u^{\prime},w^{\prime}+x^{\prime},y^{\prime},z^{\prime}\rangle, formed by the reverse operation where the sum w+xw^{\prime}+x^{\prime} is taken and uu^{\prime} is added, is included in PP. Here, we have v2(w+x)<v2(y)=v2(z)v_{2}(w^{\prime}+x^{\prime})<v_{2}(y^{\prime})=v_{2}(z^{\prime}).

      • If v2(u)>v2(y)v_{2}(u^{\prime})>v_{2}(y^{\prime}), it corresponds to (\triangle).

      • If v2(u)v2(y)v_{2}(u^{\prime})\leq v_{2}(y^{\prime}) and v2(u)v2(w+x)v_{2}(u^{\prime})\neq v_{2}(w^{\prime}+x^{\prime}), then by (3\diamondsuit_{3}), it cannot satisfy (2A) or (3A).

      • If v2(u)=v2(w+x)v_{2}(u^{\prime})=v_{2}(w^{\prime}+x^{\prime}), it corresponds to (\triangle).

      Therefore, we can conclude that u,w+x,y,zP\langle u^{\prime},w^{\prime}+x^{\prime},y^{\prime},z^{\prime}\rangle\not\in P.

  4. d).

    In case w,x,y,zP4\langle w^{\prime},x^{\prime},y^{\prime},z^{\prime}\rangle\in P_{4}:
    Suppose v2(w)<v2(x)<v2(y)<v2(z)v_{2}(w^{\prime})<v_{2}(x^{\prime})<v_{2}(y^{\prime})<v_{2}(z^{\prime}). Let c=v2(y),d=v2(z)c=v_{2}(y^{\prime}),d=v_{2}(z^{\prime}), and we will use the following facts:
    (1\heartsuit_{1}) From (4A), we have Id+1(w)=Id+1(x)=Id+1(y)=0I_{d+1}(w^{\prime})=I_{d+1}(x^{\prime})=I_{d+1}(y^{\prime})=0.
    (2\heartsuit_{2}) From (4C), we have Ic+1(w)=Ic+1(x)=1I_{c+1}(w^{\prime})=I_{c+1}(x^{\prime})=1.
    (3\heartsuit_{3}) From (4B), we have Ij(w)+Ij(x)+Ij(y)2I_{j}(w)+I_{j}(x)+I_{j}(y)\geq 2 for c+2jdc+2\leq j\leq d.
    (4\heartsuit_{4}) From (4B) and by using similar arguments as Proposition 3.3,
       we have Id+1(w+x)=Id+1(w+y)=Id+1(x+y)=1I_{d+1}(w^{\prime}+x^{\prime})=I_{d+1}(w^{\prime}+y^{\prime})=I_{d+1}(x^{\prime}+y^{\prime})=1.

    • We consider whether u,w,x,y+z\langle u^{\prime},w^{\prime},x^{\prime},y^{\prime}+z^{\prime}\rangle, formed by the reverse operation where the sum y+zy^{\prime}+z^{\prime} is taken and uu^{\prime} is added, is included in PP. Here, we have v2(w)<v2(x)<v2(y+z)<dv_{2}(w^{\prime})<v_{2}(x^{\prime})<v_{2}(y^{\prime}+z^{\prime})<d.

      • If v2(u)>dv_{2}(u^{\prime})>d, then by (1\heartsuit_{1}), it cannot satisfy (4B) or (5C).

      • If v2(u)=dv_{2}(u^{\prime})=d, then by (1\heartsuit_{1}), it cannot satisfy (4A) or (5B).

      • If v2(u)<dv_{2}(u^{\prime})<d, v2(u)v2(y+z)v_{2}(u^{\prime})\neq v_{2}(y^{\prime}+z^{\prime}), v2(u)v2(x)v_{2}(u^{\prime})\neq v_{2}(x^{\prime}) and v2(u)v2(w)v_{2}(u^{\prime})\neq v_{2}(w^{\prime}), then by (1\heartsuit_{1}) and (3\heartsuit_{3}), it cannot satisfy (4A) or (5A).

      • If v2(u)=v2(y+z)v_{2}(u^{\prime})=v_{2}(y^{\prime}+z^{\prime}), then by (2\heartsuit_{2}), it cannot satisfy (3A).

      • If v2(u)=v2(x)v_{2}(u^{\prime})=v_{2}(x^{\prime}) or v2(u)=v2(w)v_{2}(u^{\prime})=v_{2}(w^{\prime}), it corresponds to (\triangle).

      Therefore, we can conclude that u,w,x,y+zP\langle u^{\prime},w^{\prime},x^{\prime},y^{\prime}+z^{\prime}\rangle\not\in P.

    • For u,w,x+z,y\langle u^{\prime},w^{\prime},x^{\prime}+z^{\prime},y^{\prime}\rangle, formed by the reverse operation with the sum x+zx^{\prime}+z^{\prime} and addition of uu^{\prime}, and for u,w+z,x,y\langle u^{\prime},w^{\prime}+z^{\prime},x^{\prime},y^{\prime}\rangle, similarly with the sum w+zw^{\prime}+z^{\prime}, it can be shown that neither is included in PP, just as in the case where the sum y+zy^{\prime}+z^{\prime} is taken and uu^{\prime} is added.

    • We consider whether u,w,x+y,z\langle u^{\prime},w^{\prime},x^{\prime}+y^{\prime},z^{\prime}\rangle, formed by the reverse operation where the sum x+yx^{\prime}+y^{\prime} is taken and uu^{\prime} is added, is included in PP. Here, we have v2(w)<v2(x+y)<v2(z)=dv_{2}(w^{\prime})<v_{2}(x^{\prime}+y^{\prime})<v_{2}(z^{\prime})=d.

      • If v2(u)>v2(z)v_{2}(u^{\prime})>v_{2}(z^{\prime}), then by (1\heartsuit_{1}), it cannot satisfy (4C) or (5D).

      • If v2(u)=v2(z)v_{2}(u^{\prime})=v_{2}(z^{\prime}), then by (4\heartsuit_{4}), it cannot satisfy (3A).

      • If v2(u)<v2(z)v_{2}(u^{\prime})<v_{2}(z^{\prime}), v2(u)v2(x+y)v_{2}(u^{\prime})\neq v_{2}(x^{\prime}+y^{\prime}) and v2(u)v2(w)v_{2}(u^{\prime})\neq v_{2}(w^{\prime}), then by (1\heartsuit_{1}) and (4\heartsuit_{4}), it cannot satisfy (4A) or (5B).

      • If v2(u)=v2(x+y)v_{2}(u^{\prime})=v_{2}(x^{\prime}+y^{\prime}) or v2(u)=v2(w)v_{2}(u^{\prime})=v_{2}(w^{\prime}), it corresponds to (\triangle).

      Therefore, we can conclude that u,w,x+y,zP\langle u^{\prime},w^{\prime},x^{\prime}+y^{\prime},z^{\prime}\rangle\not\in P.

    • For u,w+y,x,z\langle u^{\prime},w^{\prime}+y^{\prime},x^{\prime},z^{\prime}\rangle, formed by the reverse operation with the sum w+yw^{\prime}+y^{\prime} and addition of uu^{\prime}, and for u,w+x,y,z\langle u^{\prime},w^{\prime}+x^{\prime},y^{\prime},z^{\prime}\rangle, similarly with the sum w+xw^{\prime}+x^{\prime}, it can be shown that neither is included in PP, just as in the case where the sum x+yx^{\prime}+y^{\prime} is taken and uu^{\prime} is added.

  5. e).

    In case w,x,y,zP5\langle w^{\prime},x^{\prime},y^{\prime},z^{\prime}\rangle\in P_{5}:
    Suppose v2(w)<v2(x)<v2(y)<v2(z)v_{2}(w^{\prime})<v_{2}(x^{\prime})<v_{2}(y^{\prime})<v_{2}(z^{\prime}). Let b=v2(x),c=v2(y),d=v2(z)b=v_{2}(x^{\prime}),c=v_{2}(y^{\prime}),d=v_{2}(z^{\prime}), and define the smallest ee such that e>de>d and
    (1\spadesuit_{1}) Ie+1(w)=Ie+1(x)=Ie+1(y)=Ie+1(z)=0I_{e+1}(w^{\prime})=I_{e+1}(x^{\prime})=I_{e+1}(y^{\prime})=I_{e+1}(z^{\prime})=0.
    We will use the following facts:
    (2\spadesuit_{2}) From (5A), we have Ii(w)+Ii(x)+Ii(y)+Ii(z)3I_{i}(w^{\prime})+I_{i}(x^{\prime})+I_{i}(y^{\prime})+I_{i}(z^{\prime})\geq 3 for
      d+2ied+2\leq i\leq e.
    (3\spadesuit_{3}) From (5B), we have Id+1(w)=Id+1(x)=Id+1(y)=1I_{d+1}(w^{\prime})=I_{d+1}(x^{\prime})=I_{d+1}(y^{\prime})=1 and
       Id+1(w+z)=Id+1(x+z)=Id+1(y+z)=0I_{d+1}(w^{\prime}+z^{\prime})=I_{d+1}(x^{\prime}+z^{\prime})=I_{d+1}(y^{\prime}+z^{\prime})=0.
    (4\spadesuit_{4}) From (5D), we have Ic+1(w)=Ic+1(x)=1I_{c+1}(w^{\prime})=I_{c+1}(x^{\prime})=1.
    (5\spadesuit_{5}) From (5C), we have Ij(w)+Ij(x)+Ij(y)2I_{j}(w)+I_{j}(x)+I_{j}(y)\geq 2 for c+2jdc+2\leq j\leq d,
      and from (5E), Ik(w)+Ik(x)1I_{k}(w)+I_{k}(x)\geq 1 for b+2kcb+2\leq k\leq c.
    (6\spadesuit_{6}) From (5A) and by using similar arguments as Proposition 3.3, regarding the sum of any two of w,x,y,zw^{\prime},x^{\prime},y^{\prime},z^{\prime}, Ie+1(w+x)==Ie+1(y+z)=1I_{e+1}(w^{\prime}+x^{\prime})=\cdots=I_{e+1}(y^{\prime}+z^{\prime})=1.

    • We consider whether u,w,x,y+z\langle u^{\prime},w^{\prime},x^{\prime},y^{\prime}+z^{\prime}\rangle, formed by the reverse operation where the sum y+zy^{\prime}+z^{\prime} is taken and uu^{\prime} is added, is included in PP. Here, we have v2(w)<v2(x)<v2(y+z)=cv_{2}(w^{\prime})<v_{2}(x^{\prime})<v_{2}(y^{\prime}+z^{\prime})=c.

      • If v2(u)>ev_{2}(u^{\prime})>e, then by (1\spadesuit_{1}), it cannot satisfy (4B).

      • If d<v2(u)ed<v_{2}(u^{\prime})\leq e, then by (2\spadesuit_{2}) it cannot satisfy (4A) , and by (1\spadesuit_{1}) and (6\spadesuit_{6}) it cannot satisfy (5A).

      • If v2(u)=dv_{2}(u^{\prime})=d, then by (3\spadesuit_{3}), it cannot satisfy (4A) and (5B).

      • If v2(u)<dv_{2}(u^{\prime})<d, v2(u)v2(y+z)v_{2}(u^{\prime})\neq v_{2}(y^{\prime}+z^{\prime}), v2(u)v2(x)v_{2}(u^{\prime})\neq v_{2}(x^{\prime}) and v2(u)v2(w)v_{2}(u^{\prime})\neq v_{2}(w^{\prime}), then by (5\spadesuit_{5}) it cannot satisfy (4A), and by (1\spadesuit_{1}) and (6\spadesuit_{6}) it cannot satisfy (5A).

      • If v2(u)=v2(y+z)v_{2}(u^{\prime})=v_{2}(y^{\prime}+z^{\prime}), then by (4\spadesuit_{4}), it cannot satisfy (3A).

      • If v2(u)=v2(x)v_{2}(u^{\prime})=v_{2}(x^{\prime}) or v2(u)=v2(w)v_{2}(u^{\prime})=v_{2}(w^{\prime}), it corresponds to (\triangle).

      Therefore we can conclude that u,w,x,y+zP\langle u^{\prime},w^{\prime},x^{\prime},y^{\prime}+z^{\prime}\rangle\not\in P.

    • For u,w,x+z,y\langle u^{\prime},w^{\prime},x^{\prime}+z^{\prime},y^{\prime}\rangle, formed by the reverse operation with the sum x+zx^{\prime}+z^{\prime} and addition of uu^{\prime}, and for u,w+z,x,y\langle u^{\prime},w^{\prime}+z^{\prime},x^{\prime},y^{\prime}\rangle, similarly with the sum w+zw^{\prime}+z^{\prime}, it can be shown that neither is included in PP, just as in the case where the sum y+zy^{\prime}+z^{\prime} is taken and uu^{\prime} is added.

    • We consider whether u,w,x+y,z\langle u^{\prime},w^{\prime},x^{\prime}+y^{\prime},z^{\prime}\rangle, formed by the reverse operation where the sum x+yx^{\prime}+y^{\prime} is taken and uu^{\prime} is added, is included in PP. Here, we have v2(w)<v2(x+y)<v2(z)=dv_{2}(w^{\prime})<v_{2}(x^{\prime}+y^{\prime})<v_{2}(z^{\prime})=d.

      • If v2(u)>ev_{2}(u^{\prime})>e, then by (1\spadesuit_{1}), it cannot satisfy (4B).

      • If v2(z)<v2(u)ev_{2}(z^{\prime})<v_{2}(u^{\prime})\leq e, then by (2\spadesuit_{2}) it cannot satisfy (4A), and by (1\spadesuit_{1}) and (6\spadesuit_{6}) it cannot satisfy (5A).

      • If v2(u)=v2(z)v_{2}(u^{\prime})=v_{2}(z^{\prime}), then by (3\spadesuit_{3}), it cannot satisfy (3A).

      • If v2(u)<v2(z)v_{2}(u^{\prime})<v_{2}(z^{\prime}), v2(u)v2(x+y)v_{2}(u^{\prime})\neq v_{2}(x^{\prime}+y^{\prime}) and v2(u)v2(w)v_{2}(u^{\prime})\neq v_{2}(w^{\prime}), then by (5\spadesuit_{5}) it cannot satisfy (4A) and by (1\spadesuit_{1}) and (6\spadesuit_{6}) it cannot satisfy (5A).

      • If v2(u)=v2(x+y)v_{2}(u^{\prime})=v_{2}(x^{\prime}+y^{\prime}) or v2(u)=v2(w)v_{2}(u^{\prime})=v_{2}(w^{\prime}), it corresponds to (\triangle).

      Therefore, we can conclude that u,w,x+y,zP\langle u^{\prime},w^{\prime},x^{\prime}+y^{\prime},z^{\prime}\rangle\not\in P.

    • For u,w+y,x,z\langle u^{\prime},w^{\prime}+y^{\prime},x^{\prime},z^{\prime}\rangle, formed by the reverse operation with the sum w+yw^{\prime}+y^{\prime} and addition of uu^{\prime}, and for u,w+x,y,z\langle u^{\prime},w^{\prime}+x^{\prime},y^{\prime},z^{\prime}\rangle, similarly with the sum w+xw^{\prime}+x^{\prime}, it can be shown that neither is included in PP, just as in the case where the sum x+yx^{\prime}+y^{\prime} is taken and uu^{\prime} is added.

Thus we completes the proof. \Box

Proof of Lemma 3.2. We explicitly show that for w,x,y,z\langle w,x,y,z\rangle not included in PP, it can be moved to some position that is included in PP. Let a=v2(w),b=v2(x),c=v2(y),d=v2(z)a=v_{2}(w),b=v_{2}(x),c=v_{2}(y),d=v_{2}(z).

  1. a).

    In case any of the following holds: a=b=c<da=b=c<d, a=b<c=da=b<c=d, a=b<c<da=b<c<d or, a<b=c<da<b=c<d:
    First we assume a=b<c<da=b<c<d. By Proposition 2.2, there exists a partition y=y+y′′y=y^{\prime}+y^{\prime\prime} such that v2(y)=v2(y′′)=av_{2}(y^{\prime})=v_{2}(y^{\prime\prime})=a (For example, we set y=y2ay^{\prime}=y-2^{a} and y′′=2ay^{\prime\prime}=2^{a}). Therefore, w,x,y,z\langle w,x,y,z\rangle can be moved to w,x,y,y′′P1\langle w,x,y^{\prime},y^{\prime\prime}\rangle\in P_{1}. Similarly, in the cases a=b=c<da=b=c<d, a=b<c=da=b<c=d and a<b=c<da<b=c<d, it can be moved to some position in P1P_{1}.

  2. b).

    In case a<b=c=da<b=c=d and (2A) is not satisfied:
    By partitioning ww into w=w2bw^{\prime}=w-2^{b} and w′′=2bw^{\prime\prime}=2^{b}, it can be moved to w,w′′,x,yP2\langle w^{\prime},w^{\prime\prime},x,y\rangle\in P_{2}.

  3. c).

    In case a<b<c=da<b<c=d and at least one of (3A),(3B), or (3C) is not satisfied:

    • Suppose that (3A) is not satisfied and Ic+1(w)=1I_{c+1}(w)=1. By partitioning ww into w=w2cw^{\prime}=w-2^{c} and w′′=2cw^{\prime\prime}=2^{c}, it can be moved to w,w′′,y,zP2\langle w^{\prime},w^{\prime\prime},y,z\rangle\in P_{2}. The same applies in the case when Ic+1(x)=1I_{c+1}(x)=1.

    • Suppose that (3C) is not satisfied and Ib+1(w)=0I_{b+1}(w)=0. By partitioning yy into y=y2by^{\prime}=y-2^{b} and y′′=2by^{\prime\prime}=2^{b}, it can be moved to w,x,y,y′′P2\langle w,x,y^{\prime},y^{\prime\prime}\rangle\in P_{2}.

    • Suppose that (3C) is satisfied but (3B) is not. For the smallest kk such that c<k<dc<k<d and Ik+1(w)=Ik+1(x)=0I_{k+1}(w)=I_{k+1}(x)=0, by partitioning yy into y=y2ky^{\prime}=y-2^{k} and y′′=2ky^{\prime\prime}=2^{k}, it can be moved to w,x,y,y′′P3\langle w,x,y^{\prime},y^{\prime\prime}\rangle\in P_{3}.

  4. d).

    In case a<b<c<da<b<c<d and (4A) is satisfied, and at least one of (4B), (4C), (4D), and (4E) is not satisfied:

    • Suppose that (4E) is not satisfied, that is Ib+1(w)=0I_{b+1}(w)=0. By partitioning yy into y=y2by^{\prime}=y-2^{b} andy′′=2by^{\prime\prime}=2^{b}, it can be moved to w,x,y,y′′P2\langle w,x,y^{\prime},y^{\prime\prime}\rangle\in P_{2}.

    • Suppose that (4E) is satisfied but (4D) is not. For the smallest kk such that b<k<cb<k<c and Ik+1(w)=Ik+1(x)=0I_{k+1}(w)=I_{k+1}(x)=0, by partitioning yy into y=y2ky^{\prime}=y-2^{k} and y′′=2ky^{\prime\prime}=2^{k}, it can be moved to w,x,y,y′′P3\langle w,x,y^{\prime},y^{\prime\prime}\rangle\in P_{3}.

    • Suppose that (4C) is not satisfied and Ic+1(w)=0I_{c+1}(w)=0. By partitioning zz into z=z2cz^{\prime}=z-2^{c} and z′′=2cz^{\prime\prime}=2^{c}, it can be moved to w,y,z,z′′P2\langle w,y,z^{\prime},z^{\prime\prime}\rangle\in P_{2}. The same applies in the case when Ic+1(x)=0I_{c+1}(x)=0.

    • Suppose (4C), (4D) and (4E) are satisfied but (4B) is not. Define the smallest kk such that c<k<dc<k<d and at least two of Ik+1(w)I_{k+1}(w), Ik+1(x)I_{k+1}(x), Ik+1(y)I_{k+1}(y) are 0. For example, if Ik+1(w)=Ik+1(x)=0I_{k+1}(w)=I_{k+1}(x)=0, then by partitioning zz into z=z2kz^{\prime}=z-2^{k} and z′′=2kz^{\prime\prime}=2^{k}, it can be moved to w,x,z,z′′P3\langle w,x,z^{\prime},z^{\prime\prime}\rangle\in P_{3}. The same applies in other cases.

  5. e).

    In case a<b<c<da<b<c<d and (4B), (4C), (4D) and (4E) are satisfied, but (4A) is not:

    Since Id+1(w)+Id+1(x)+Id+1(y){1,2}I_{d+1}(w)+I_{d+1}(x)+I_{d+1}(y)\in\{1,2\}, at least one of Id+1(w)I_{d+1}(w), Id+1(x)I_{d+1}(x) or Id+1(y)I_{d+1}(y) is 1, and at least one of them is 0. For example, if Id+1(w)=1I_{d+1}(w)=1 and Id+1(x)=0I_{d+1}(x)=0, then by partitioning ww into w=w2dw^{\prime}=w-2^{d} and w′′=2dw^{\prime\prime}=2^{d}, it can be moved to w,x,w′′,zP3\langle w^{\prime},x,w^{\prime\prime},z\rangle\in P_{3}. The same applies in other cases.

  6. f).

    In case a<b<c<da<b<c<d and (5B) is satisfied, and at least one of (5C), (5D), (5E) and (5F) is not satisfied:

    It can be shown similarly as in case d) above.

  7. g).

    In case a<b<c<da<b<c<d and (5C), (5D), (5E), (5F) are satisfied, but (5B) is not:

    It can be shown similarly as in case e) above.

  8. h).

    In case a<b<c<da<b<c<d and (5B), (5C), (5D), (5E), (5F) are satisfied, but (5A) is not:

    Define the smallest kk such that k>dk>d and Ik+1(w)+Ik+1(x)+Ik+1(y)+Ik+1(z){1,2}I_{k+1}(w)+I_{k+1}(x)+I_{k+1}(y)+I_{k+1}(z)\in\{1,2\}. Since at least one of Ik+1(w)I_{k+1}(w), Ik+1(x)I_{k+1}(x), Ik+1(y)I_{k+1}(y) and Ik+1(z)I_{k+1}(z) is 1, and at least two of them are 0. For example, if Ik+1(w)=1I_{k+1}(w)=1 and Ik+1(x)=Ik+1(y)=0I_{k+1}(x)=I_{k+1}(y)=0, then by partitioning ww into w=w2kw^{\prime}=w-2^{k} and w′′=2kw^{\prime\prime}=2^{k}, it can be moved to w,x,y,w′′P4\langle w^{\prime},x,y,w^{\prime\prime}\rangle\in P_{4}. The same applies in other cases.

Thus we completes the proof. \Box

4 Single-delete Nim with five or more piles

As discussed earlier, the winning and losing conditions for Single-delete Nim with four piles are significantly more complex than those for three or fewer piles. Additionally, it is not yet possible to even conjecture the conditions for scenarios with more piles. For Single-delete Nim with five or more piles, this paper is limited to presenting the following two almost self-evident propositions. In particular, Proposition 4.2 suggests that, unlike the conditions in Theorem 1.4 and Theorem 1.5(1), even if the number of times each pile’s stone count can be divided by 2 is the same, the position does not necessarily become a P-position when there are more piles, indicating that the winning and losing conditions become even more complex.

Proposition 4.1.

In Single-delete Nim with n2n\geq 2 piles, a position x1,x2,\langle x_{1},x_{2},
x3,,xnx_{3},\ldots,x_{n}\rangle is a P-position if all xix_{i} are odd.

Proof. Let P0P_{0} be the set of positions where all xix_{i} are odd. When the player on their turn makes a move from x1,x2,x3,,xnP0\langle x_{1},x_{2},x_{3},\ldots,x_{n}\rangle\in P_{0} to x1,x2,x3,,xn\langle x^{\prime}_{1},x^{\prime}_{2},x^{\prime}_{3},\ldots,x^{\prime}_{n}\rangle, exactly one of x1,x2,x3,,xnx^{\prime}_{1},x^{\prime}_{2},x^{\prime}_{3},\ldots,x^{\prime}_{n} becomes even. In the next move, the opposite player eliminates one of the odd piles and splits the even pile into two odd piles, thus returning the position back to P0P_{0}. Therefore, at the position x1,x2,x3,,xnP0\langle x_{1},x_{2},x_{3},\ldots,x_{n}\rangle\in P_{0}, the player on their turn is always forced to take their turn in a position belonging to P0P_{0}, eventually reaching the terminal position 1,1,1,,1\langle 1,1,1,\ldots,1\rangle. \Box

Proposition 4.2.

In Single-delete Nim with n2n\geq 2 piles, if the remainder when nn is divided by 3 is 2, then 2,2,2,,2\langle 2,2,2,\ldots,2\rangle is a N-position. For any other value of nn, 2,2,2,,2\langle 2,2,2,\ldots,2\rangle is a P-position.

Proof. In Single-delete Nim, starting from the initial position 2,2,2,\langle 2,2,2,
,2\ldots,2\rangle, the positions that appear during the game can be described as having kk piles with 2 stones and nkn-k piles with one stone ?i1kn1\leq k\leq n?j. Considering that on each turn, the player can only remove a pile with 1 or 2 stones and split one of the piles with 2 stones, the game can be reduced to an equivalent stone-picking game of the following form by focusing on the value of kk.

  • On the first turn, the player takes 2 stones from the total of nn stones. On subsequent turns, each player may take either 1 or 2 stones.

  • The player who takes the last stone wins the game.

Therefore, when the remainder of nn divided by 3 is 2, the first player can always win by taking 2 stones initially. If the opponent then takes 2 stones, the first player should take 1 stone on the next turn; if the opponent takes 1 stone, the first player should take 2 stones. This strategy guarantees a win. The proof for the case when the remainder of nn divided by 3 is not 2 is omitted. \Box

Acknowledgements

We would like to express our gratitude to Dr. Koki Suetsugu, Prof. Tomoaki Abuku, and Prof. Ko Sakai for their invaluable comments regarding the study presented in this paper. This work was supported by JSPS KAKENHI Grant Number JP21K12191.

References

  • [1] Abuku, T., Sakai, K., Shinoda, M., Suetsugu, K.: Some extensions of delete nim. arXiv preprint arXiv(2301.12964)) (2023), https://arxiv.org/abs/2301.12964
  • [2] Abuku, T., Suetsugu, K.: Delete Nim. J. Math. Tokushima Univ. 55, 75–81 (2021)
  • [3] Bouton, C.L.: Nim, a game with a complete mathematical theory. Ann. of Math. 3(1-4), 35–39 (1901/02)
  • [4] Moore, E.H.: A generalization of the game called nim. Ann. of Math. 11(3), 93–94 (1910)
  • [5] Sakai, K.: Dividing Pizza with Mathematics (“Sugaku de Pizza wo Kiriwakeru” in Japanese). Alice’s Adventure in Puzzle-Land 4, Nikkei Science Inc. (2021)
  • [6] Stankova, Z., Rike, T. (eds.): A Decade of the Berkeley Math Circle: The American Experience, Volume I, MSRI Mathematical Circles Library, vol. 14. American Mathematical Society (2008)
  • [7] Wythoff, W.A.: A modification of the game of Nim. Nieuw Arch. Wiskd. 7(2), 199–202 (1907)